CN112037173A - Chromosome detection method and device and electronic equipment - Google Patents
Chromosome detection method and device and electronic equipment Download PDFInfo
- Publication number
- CN112037173A CN112037173A CN202010774367.4A CN202010774367A CN112037173A CN 112037173 A CN112037173 A CN 112037173A CN 202010774367 A CN202010774367 A CN 202010774367A CN 112037173 A CN112037173 A CN 112037173A
- Authority
- CN
- China
- Prior art keywords
- chromosome
- sub
- images
- image
- feature extraction
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 210000000349 chromosome Anatomy 0.000 title claims abstract description 333
- 238000001514 detection method Methods 0.000 title claims abstract description 132
- 238000000605 extraction Methods 0.000 claims abstract description 70
- 238000000034 method Methods 0.000 claims description 36
- 238000004590 computer program Methods 0.000 claims description 12
- 230000000877 morphologic effect Effects 0.000 claims description 12
- 238000012545 processing Methods 0.000 claims description 8
- 230000011218 segmentation Effects 0.000 claims description 5
- 238000012549 training Methods 0.000 description 18
- 230000008569 process Effects 0.000 description 13
- 238000013528 artificial neural network Methods 0.000 description 6
- 238000004422 calculation algorithm Methods 0.000 description 5
- 238000007477 logistic regression Methods 0.000 description 5
- 239000002131 composite material Substances 0.000 description 4
- 238000010586 diagram Methods 0.000 description 4
- 230000000694 effects Effects 0.000 description 3
- 230000005484 gravity Effects 0.000 description 3
- 108091060290 Chromatid Proteins 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 2
- 210000004756 chromatid Anatomy 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 239000011159 matrix material Substances 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 241001465754 Metazoa Species 0.000 description 1
- 230000004913 activation Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 210000002230 centromere Anatomy 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 201000010099 disease Diseases 0.000 description 1
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 description 1
- 238000004043 dyeing Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 210000003917 human chromosome Anatomy 0.000 description 1
- 238000003709 image segmentation Methods 0.000 description 1
- 239000000178 monomer Substances 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 230000035479 physiological effects, processes and functions Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0012—Biomedical image inspection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30004—Biomedical image processing
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Data Mining & Analysis (AREA)
- Radiology & Medical Imaging (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Life Sciences & Earth Sciences (AREA)
- Medical Informatics (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Quality & Reliability (AREA)
- Evolutionary Computation (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Investigating Or Analysing Biological Materials (AREA)
- Measuring Or Testing Involving Enzymes Or Micro-Organisms (AREA)
Abstract
The embodiment of the invention provides a chromosome detection method, a chromosome detection device and electronic equipment, wherein the chromosome detection method comprises the following steps: dividing the acquired chromosome image into N sub-images, wherein N is an integer greater than 1; inputting the N sub-images into a chromosome detection model to obtain a chromosome detection result, wherein the chromosome detection model comprises a prediction module and N feature extraction modules, the N feature extraction modules correspond to the N sub-images one by one, the input ends of the N feature extraction modules are used for receiving the N sub-images, the output ends of the N feature extraction modules are connected to the input end of the prediction module, and the output end of the prediction module is used for outputting the chromosome detection result. The embodiment of the invention can reduce the over-fitting problem of the chromosome detection model and improve the robustness of the chromosome detection model and the accuracy of chromosome detection.
Description
Technical Field
The invention relates to the technical field of chromosome detection, in particular to a chromosome detection method, a chromosome detection device and electronic equipment.
Background
As is well known, chromosome detection is of great significance for disease detection, prediction, research and the like; with the development of information technology, the detection of chromosomes through a neural network has been applied to a certain extent.
Because the number of the features included in the chromosome is large and the repeatability among the features is large, the karyotypes among the non-homologous chromosomes, such as the length, the banding and the like, have large differences, so that the neural network based on the mixed training of different chromatids adopted in the prior art is often over-fitted, and the chromosome detection effect is poor.
Disclosure of Invention
The embodiment of the invention provides a chromosome detection method, a chromosome detection device and electronic equipment, and aims to solve the problems that in the prior art, a neural network based on mixed training of different chromatids is often over-fitted and the chromosome detection effect is poor.
In order to solve the technical problem, the invention is realized as follows:
in a first aspect, an embodiment of the present invention provides a chromosome detection method, including:
dividing the acquired chromosome image into N sub-images, wherein N is an integer greater than 1;
inputting the N sub-images into a chromosome detection model to obtain a chromosome detection result, wherein the chromosome detection model comprises a prediction module and N feature extraction modules, the N feature extraction modules correspond to the N sub-images one by one, the input ends of the N feature extraction modules are used for receiving the N sub-images, the output ends of the N feature extraction modules are connected to the input end of the prediction module, and the output end of the prediction module is used for outputting the chromosome detection result.
In a second aspect, an embodiment of the present invention further provides a chromosome detection apparatus, including:
the segmentation module is used for segmenting the acquired chromosome image into N sub-images, wherein N is an integer greater than 1;
the detection module is used for inputting the N sub-images into a chromosome detection model to obtain a chromosome detection result, wherein the chromosome detection model comprises a prediction module and N feature extraction modules, the N feature extraction modules correspond to the N sub-images one by one, the input ends of the N feature extraction modules are used for receiving the N sub-images, the output ends of the N feature extraction modules are connected to the input end of the prediction module, and the output end of the prediction module is used for outputting the chromosome detection result.
In a third aspect, an embodiment of the present invention further provides an electronic device, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor implements the above method when executing the computer program.
In a fourth aspect, the present invention also provides a computer-readable storage medium, where a computer program is stored, and when the computer program is executed by a processor, the computer program implements the method described above.
The chromosome detection method provided by the embodiment of the invention divides the acquired chromosome image into N sub-images, inputs the N sub-images into the chromosome detection model, and the chromosome detection module is provided with N feature extraction modules for correspondingly receiving the N sub-images and a prediction module for receiving the output of the N feature extraction modules, and the prediction module can further output the chromosome detection result. Compared with the mode that the whole chromosome is input into the neural network for detection in the prior art, the chromosome image is divided into N sub-images, focusing can be carried out on all parts of the chromosome, attention to all parts of the chromosome is improved, the feature utilization rate is improved, and meanwhile, the number of training samples can be increased in the process of training to obtain a chromosome detection model; and then the problem of over-fitting of the chromosome detection model can be reduced, and the robustness of the chromosome detection model and the chromosome detection accuracy are improved.
Drawings
FIG. 1 is a flowchart of a chromosome detection method according to an embodiment of the present invention;
FIG. 2 is a flowchart of obtaining chromosome detection results according to an embodiment of the present invention;
FIG. 3 is a network framework diagram of a chromosome detection model in an embodiment of the invention;
FIG. 4 is a diagram illustrating initial homologous chromosome pairing results in a specific application scenario;
FIG. 5 is a diagram illustrating an initial homologous chromosome pairing result after adjustment in a specific application scenario;
fig. 6 is a schematic structural diagram of a chromosome detection apparatus according to an embodiment of the present invention.
Detailed Description
In order to make the technical problems, technical solutions and advantages of the present invention more apparent, the following detailed description is given with reference to the accompanying drawings and specific embodiments. In the following description, specific details such as specific configurations and components are provided only to help the full understanding of the embodiments of the present invention. Thus, it will be apparent to those skilled in the art that various changes and modifications may be made to the embodiments described herein without departing from the scope and spirit of the invention. In addition, descriptions of well-known functions and constructions are omitted for clarity and conciseness.
Unless defined otherwise, technical or scientific terms used herein shall have the ordinary meaning as understood by one of ordinary skill in the art to which this invention belongs. The use of "first," "second," and similar terms in the present application do not denote any order, quantity, or importance, but rather the terms are used to distinguish one element from another. Also, the use of the terms "a" or "an" and the like do not denote a limitation of quantity, but rather denote the presence of at least one.
As shown in fig. 1, the chromosome detection method provided by the embodiment of the present invention includes:
102, inputting the N sub-images into a chromosome detection model to obtain a chromosome detection result, wherein the chromosome detection model comprises a prediction module and N feature extraction modules, the N feature extraction modules correspond to the N sub-images one by one, the input ends of the N feature extraction modules are used for receiving the N sub-images, the output ends of the N feature extraction modules are all connected to the input end of the prediction module, and the output end of the prediction module is used for outputting the chromosome detection result.
It is easily understood that the chromosome image may refer to an image of a certain chromosome to be detected; the method of acquiring the chromosome image is not particularly limited, and for example, the chromosome image may be obtained by performing processes such as chromosome contour recognition and image segmentation on the captured original image.
The chromosome image is divided into N sub-images, which can be regarded as dividing the chromosome in the chromosome image into N segments, and each sub-image can have data of a corresponding segment of chromosome. For example, a chromosome may be divided into N parts, each part corresponding to a sub-image, according to skeleton lines of the chromosome in the chromosome image; of course, in some possible embodiments, the chromosome image may be divided into N sub-images according to information such as pixel coordinates of the chromosome image. That is, the segmentation mode for the chromosome image can be determined according to actual needs, and is not particularly limited here.
In this embodiment, the chromosome detection model may be obtained by training the constructed original chromosome detection neural network through a training sample. The chromosome detection module comprises a prediction module and N feature extraction modules, wherein the input end of each feature extraction module can receive the sub-images, the input end of the prediction module can also receive the contents output by the N feature extraction modules, and the output end of the prediction module can output the chromosome detection result. The chromosome detection result herein may refer to karyotype, polarity, and the like of the chromosome, and will be specifically described below.
The network frames and the network parameters corresponding to the N feature extraction modules may be the same or different, and the features of the input sub-images may be extracted. It is easy to understand that, in this embodiment, the input of the feature extraction module is not an image of the whole chromosome, but is based on the sub-images obtained by segmenting the chromosome image, so on one hand, different parts of the chromosome can be focused, and the attention and feature utilization rate of each part of the chromosome can be improved, and on the other hand, any sample chromosome image can be segmented into a plurality of sample sub-images in the process of obtaining the chromosome detection model by training, so as to increase the number of training samples; and then the problem of over-fitting of the chromosome detection model can be reduced, and the robustness of the chromosome detection model is improved.
The chromosome detection method provided by the embodiment of the invention divides the acquired chromosome image into N sub-images, inputs the N sub-images into the chromosome detection model, and the chromosome detection module is provided with N feature extraction modules for correspondingly receiving the N sub-images and a prediction module for receiving the output of the N feature extraction modules, and the prediction module can further output the chromosome detection result. Compared with the mode that the whole chromosome is input into the neural network for detection in the prior art, the chromosome image is divided into N sub-images, focusing can be carried out on all parts of the chromosome, attention to all parts of the chromosome is improved, the feature utilization rate is improved, and meanwhile, the number of training samples can be increased in the process of training to obtain a chromosome detection model; and then the problem of over-fitting of the chromosome detection model can be reduced, and the robustness of the chromosome detection model and the chromosome detection accuracy are improved.
In one example, the following method may be specifically adopted for the segmentation of the chromosome image: and processing the chromosome image based on a thinning algorithm to obtain a skeleton line, and cutting the chromosome into three parts based on trisection on the skeleton line, wherein each part corresponds to one sub-image.
The thinning algorithm generally refers to a skeletonization algorithm of a binary image, and a skeleton line is obtained by processing a chromosome image based on the thinning algorithm, belonging to the conventional application of the thinning algorithm in the field of chromosome detection and not being described again here.
It should be emphasized that the trisection division shown in this example is only an example of one possible chromosome image division, and in practical applications, the chromosome may be divided into other numbers of halves, or the division is not performed in the manner of the trisection, but performed in a preset length ratio, and the like. In other words, the dividing mode of the chromosome image can be selected according to actual needs.
Optionally, the prediction module comprises a karyotype prediction unit, and the chromosome detection result comprises a chromosome karyotype;
the step 102 of inputting the N sub-images into a chromosome detection model to obtain a chromosome detection result includes:
inputting the N sub-image inputs into the N feature extraction modules correspondingly to obtain N feature sets;
and inputting the N feature sets into the karyotype prediction unit to obtain the chromosome karyotype.
The karyotype may refer to the length, streak, or the like of the chromosome. In this embodiment, each feature extraction module may perform feature extraction on the corresponding sub-image to obtain a corresponding feature set, and the karyotype prediction unit may receive the N feature sets output by the N feature extraction modules, and process the N feature sets by, for example, logistic regression, to obtain the chromosome karyotype.
In this embodiment, the karyotype of the chromosome is detected, which is helpful for further implementing operations such as chromosome number judgment, homologous chromosome pairing, and the like.
Optionally, to further improve the detection accuracy of the chromosome karyotype, the inputting the N feature sets into the karyotype prediction unit to obtain the chromosome karyotype includes:
predicting each feature set in the N feature sets by the karyotype prediction unit to obtain N initial prediction results;
and calculating by the karyotype prediction unit according to the N initial karyotype prediction results and the preset weight of each initial karyotype prediction result to obtain the chromosome karyotype.
This embodiment may be considered as defining a preferred mode of operation of the karyotype prediction unit. Specifically, the karyotype predicting unit may perform karyotype prediction on each feature set to obtain N initial prediction results, and perform composite determination on the N initial prediction results to obtain a chromosome karyotype for final output.
For example, if N is 3, the three feature sets are respectively denoted as a feature set a, a feature set B, and a feature set C, and are respectively corresponding to outputs of the three feature extraction modules; the karyotype is represented by chromosome numbering.
The three initial prediction results obtained by the karyotype prediction unit respectively predicting the three feature sets are as follows:
wherein,
representing the probability value of the chromosome being chromosome 1 predicted according to the feature set A,
representing the probability value of the chromosome being chromosome 1 predicted according to the feature set B,
the probability value of the chromosome being chromosome 1 is obtained by predicting the feature set C.
Each probability value can correspond to a preset weight; of course, in practical applications, the preset weight may be a weight given to different feature extraction modules. For example, if the preset weights of the three probability values are 0.3, 0.4, and 0.3, respectively, the result of the karyotype prediction unit calculated for the initial prediction result and the preset weights is: the probability that chromosome 1 is 0.3 × 0.9+0.4 × 0.95+0.3 × 0.8 is 0.89; further, if there is a probability threshold of 0.8, since the probability value of 0.89 obtained by the weight calculation is greater than the probability threshold, it can be determined that the chromosome is chromosome 1, that is, the chromosome karyotype described above.
Of course, the above is only an example of the working process of the karyotype prediction unit, and the embodiment of the chromosome karyotype in practical application may also be length, specific gravity of the banding pattern, etc.; the preset weight can be set according to actual needs, and the like.
Optionally, the prediction module further comprises a polarity prediction unit, and the chromosome detection result further comprises a chromosome polarity;
after the N sub-image inputs are respectively and correspondingly input to the N feature extraction modules to obtain N feature sets, the method further includes:
and inputting the N feature sets into the polarity prediction unit to obtain the polarity of the chromosome.
The polarity of the chromosome may generally refer to the polarity in terms of morphology or physiology, and for example, in the morphology, the chromosome may have a long arm and a short arm in the direction of longitudinal extension based on the position of the centromere.
In this embodiment, the chromosome polarity is obtained by prediction, and the specific extending direction of the chromosome, for example, whether the short arm is above and the long arm is below, can be obtained; thereby being beneficial to carrying out accurate comparison of characteristics among different dyeing monomers and avoiding comparison errors caused by different comparison starting ends; besides, the method is also helpful to ensure the consistency of the extending directions of the chromosomes when the chromosomes are displayed, and is convenient for a user to observe and correct.
Optionally, each sub-image in the N sub-images has a corresponding image sequence number;
inputting the N feature sets into the polarity prediction unit to obtain the polarity of the chromosome, wherein the method comprises the following steps:
merging the N feature sets according to the image sequence number of the sub-image corresponding to each feature set through the polarity prediction unit to obtain a merging result;
and predicting by the polarity prediction unit based on the combined result to obtain the chromosome polarity.
In this embodiment, considering that N sub-images generally have a fixed arrangement order, assigning corresponding image sequence numbers is helpful for determining the merging order of N feature sets when merging the N feature sets subsequently; further, when the polarity prediction unit predicts the merged result, the obtained chromosome polarity can also be matched with the specific extending direction of the chromosome in the chromosome image.
Certainly, in practical application, the N feature extraction modules may also have corresponding serial numbers, and the N sub-images may be matched and input to the corresponding feature extraction modules according to the serial numbers of the sub-images and the serial numbers of the feature extraction modules; when the subsequent polarity prediction unit merges the N feature sets, the merging order of the N feature sets can be determined directly according to the serial numbers of the feature extraction modules outputting the feature sets.
Optionally, as shown in fig. 2, the step 102 of inputting the N sub-images into a chromosome detection model to obtain a chromosome detection result includes:
In this embodiment, a specific image or object may be used to block part or all of the image to be processed, so as to control the image processing area or process; the specific image or object due to occlusion is referred to as a mask, i.e., corresponds to the first mask described above. In this embodiment, the first mask may be generated randomly, and the first mask may be used to extract the effective region from the first convolution image to obtain the first processed image.
In the embodiment of the invention, a mask processing process is added in the feature extraction module; it is easy to understand that the feature extraction module, as a component of the chromosome detection model, is obtained by training based on a training sample; in the process of obtaining the chromosome detection model by training the training samples, due to the addition of the random mask, the repeatability between the training samples is reduced, the quality of the training sample set is improved, the over-fitting condition of the chromosome detection model can be effectively avoided, and the robustness of the chromosome detection model is improved.
As shown in fig. 3, in one example, each feature extraction module includes a convolution layer, a mask Unit (Partial Unit), a non-saturation activation function layer (leakage Relu), and a multi-scale mean value feature extraction layer (multilevemealuefects). The convolutional layer may be 32 convolutional networks with a size of 3 × 3 and a filling width of 1 step of 1 (denoted as Cnov 3 × 3padding — 1stride — 1), the Partial Unit enhances the robustness of the chromosome detection model by randomly generating an effective region mask, and the calculation process of the Partial Unit may be represented as:
wherein, x'(i,j)The value of a pixel point (hereinafter referred to as a first pixel point) with coordinates (i, j) in the first processed image; w is a network parameter in the chromosome detection model, the network parameter is usually adjusted in real time in the process of training the chromosome detection model, and the network parameter can be regarded as a fixed value after the training is finished; x(i,j)A pixel matrix associated with a first pixel point in the first convolution image; m(i,j)A matrix of masks associated with the first pixel;
to solve for the tensor product, | | | | is the operation of solving for the norm, the specific solving method is not limited here, and can be regarded as a normalization process, r(i,j)Is the specific gravity value associated with the first pixel point.
And the first processed image input by the mask unit passes through an active layer of Leaky Relu once, and is input into the prediction module after the multi-scale mean value characteristic is extracted, so as to obtain a chromosome detection result.
To facilitate understanding of the working principle of the whole chromosome detection model, the present example is continued with the prediction module mentioned in the above embodiments.
Referring also to FIG. 3, the prediction module includes a kernel type prediction unit and a polarity prediction unit.
Wherein, the karyotype prediction unit includes: 3 fully connected & logistic regression layers (FC & Sofrmax) and composite fault layers (composite); the 3 full-connection and logistic regression layers are used for respectively generating 3 initial prediction results, and the composite fault judgment layer combines the initial prediction results according to preset weight to output chromosome karyotypes (types);
the polarity prediction unit comprises a merging layer (concat) and a full-connection & logistic regression layer (FC & Sofrmax), wherein the merging layer is used for merging the 3 groups of multi-scale mean features, inputting merging results into the full-connection & logistic regression layer and finally outputting chromosome polarity (vertical).
Optionally, after the inputting the N sub-images into the chromosome detection model and obtaining the chromosome detection result, the method further includes:
generating an initial homologous chromosome pairing result according to the chromosome detection result, wherein the chromosome detection result comprises morphological information of chromosomes, and the initial homologous chromosome pairing result comprises an initial number of each chromosome;
respectively acquiring first prior data corresponding to each chromosome, and inputting the first chromosome into an alternative set under the condition that the first chromosome with form information not matched with the corresponding first prior data exists, wherein the first prior data is prior data matched with the initial number of the corresponding chromosome;
respectively obtaining at least one second prior data corresponding to each second chromosome, and updating an initial number of a third chromosome to a number corresponding to any second prior data when the third chromosome exists in all the second chromosomes, wherein the morphological information of the third chromosome is matched with any second prior data in the corresponding at least one second prior data, the second chromosome is a chromosome in all chromosomes in the chromosome detection results except the first chromosome, and the second prior data is prior data matched with a number adjacent to the initial number of the corresponding chromosome;
determining all target numbers with the number of corresponding chromosomes being less than 2 from the updated initial numbers, obtaining third prior data corresponding to each target number, and determining the number of the first chromosome in the alternative set according to the third prior data.
In a specific application scenario, for a human chromosome, the initial homologous chromosome pairing result may be the pairing result shown in fig. 4, as can be seen from fig. 4, the initial homologous chromosome pairing result directly generated based on the chromosome detection result is not accurate enough, and in the prior art, the pairing result is usually modified by means of human intervention.
In this embodiment, the initial homologous chromosome pairing result is modified using the prior knowledge, that is, the various types of prior data mentioned above, so that the accuracy of the finally obtained homologous chromosome pairing result can be effectively improved.
Specifically, in the initial homologous chromosome matching result, each chromosome is substantially assigned with a corresponding initial number (each number may correspond to a karyotype), and the matching result is to integrate chromosomes with the same initial numbers; the basis for determining the initial number may be morphological information of the chromosome (referred to as first morphological information) in the chromosome detection result, such as the length of the chromosome, the specific gravity of the banding pattern, and the high latitude feature output by the chromosome detection model. Similarly, the a priori data may also include the morphological information (denoted as second morphological information) and associate the chromosome number with the second morphological information.
Based on the above description, it can be considered that the present embodiment performs three chromosome traversals for the initial homologous chromosome pairing result, specifically:
in the first traversal, according to the initial number of a certain chromosome, determining second form information (corresponding to first prior data at this time) of the chromosome corresponding to the initial number, then comparing the first form information of the chromosome with the corresponding second form information, if the difference displayed by the comparison result is smaller than a threshold value, considering that the initial number of the chromosome is correct, and if the difference displayed by the comparison result is larger than the threshold value, considering that the initial number of the chromosome is wrong, and inputting the chromosome (i.e. the first chromosome) into an alternative set;
the second traversal is performed mainly for chromosomes outside the candidate set, and if a corresponding initial number is 3, the first form information of the chromosome may be compared with the second form information of the adjacent chromosome 2 and chromosome 4 (which corresponds to the second prior data) respectively, to determine whether there is an erroneously visited chromosome, and if so, for example, if the first form information of the chromosome is matched with the second form information of the chromosome 2, the number of the chromosome is updated to be 2. Of course, the determination of the adjacent numbers can be determined according to actual needs, for example, for chromosome 3, the adjacent numbers to be determined here may be number 2 and number 4 described above, may be number 2 alone or number 4 alone, or may be number 1, number 2, number 4, and number 5.
Through the first two times of traversal, the condition that more than three chromosomes exist under a certain chromosome number can be basically eliminated, but the condition that less than 2 chromosomes exist under the certain chromosome number possibly exists; therefore, in the third traversal, all target numbers corresponding to chromosomes with the number less than 2 are determined, and the chromosomes with the first form information meeting the second form information in the place are screened from the alternative set by using the second form information (corresponding to the third prior data at this time) of the chromosomes corresponding to the target numbers, and are endowed with new numbers.
Through the processing process, the corresponding chromosome number can be determined for each chromosome more accurately, namely the karyotype of each chromosome is judged more accurately, and the accurate pairing and calibration of homologous chromosomes can be realized according to the karyotype of each chromosome; the final pairing and calibration are combined as shown in fig. 5, and it can be seen that the condition that more than 3 chromosomes exist under a certain chromosome number is eliminated, and the calibration result is more accurate and reasonable.
Of course, it will be appreciated that the calibration of human homologous chromosomes is shown in FIGS. 4 and 5, for a total of 23 pairs of chromosomes; the chromosome detection method provided by the above embodiment can also be applied to other animals and plants, and the logarithms of homologous chromosomes obtained correspondingly can be changed correspondingly.
The embodiment of the invention also provides a chromosome detection device, which comprises:
a segmentation module 601, configured to segment the acquired chromosome image into N sub-images, where N is an integer greater than 1;
the detecting module 602 is configured to input the N sub-images into a chromosome detection model to obtain a chromosome detection result, where the chromosome detection model includes a predicting module and N feature extracting modules, the N feature extracting modules correspond to the N sub-images one to one, an input end of each of the N feature extracting modules is configured to receive the N sub-images, output ends of the N feature extracting modules are all connected to an input end of the predicting module, and an output end of the predicting module is configured to output the chromosome detection result.
Optionally, the prediction module comprises a karyotype prediction unit, and the chromosome detection result comprises a chromosome karyotype;
the detection module 602 includes:
the first acquisition unit is used for correspondingly inputting the N sub-image inputs into the N characteristic extraction modules respectively to obtain N characteristic sets;
and the second acquisition unit is used for inputting the N feature sets into the karyotype prediction unit to obtain the chromosome karyotype.
Optionally, the second obtaining unit includes:
the first obtaining subunit is configured to predict, by the karyotype prediction unit, each feature set in the N feature sets, respectively to obtain N initial prediction results;
and the second obtaining subunit is used for calculating according to the N initial karyotype prediction results and the preset weight of each initial karyotype prediction result through the karyotype prediction unit to obtain the chromosome karyotype.
Optionally, the prediction module further comprises a polarity prediction unit, and the chromosome detection result further comprises a chromosome polarity;
the detection module 602 further includes:
and the third acquisition unit is used for inputting the N feature sets into the polarity prediction unit to obtain the polarity of the chromosome.
Optionally, each sub-image in the N sub-images has a corresponding image sequence number;
the third obtaining unit includes:
the third obtaining subunit is configured to merge, by the polarity prediction unit, the N feature sets according to the image sequence number of the sub-image corresponding to each feature set, so as to obtain a merging result;
and the fourth obtaining subunit is configured to perform prediction by the polarity prediction unit based on the merged result to obtain the chromosome polarity.
Optionally, the detecting module 602 includes:
a fourth obtaining unit, configured to perform convolution processing on a first sub-image through a first feature extraction module to obtain a first convolution image, where the first feature extraction module is any one of the N feature extraction modules, and the first sub-image is a sub-image corresponding to the first feature extraction module in the N sub-images;
a generating unit, configured to randomly generate a first mask through the first feature extraction module, and generate a first processed image according to the first volume image and the first mask;
the fifth acquisition unit is used for performing feature extraction on the basis of the first processed image through the first feature extraction module to obtain a first feature set;
and the sixth acquisition unit is used for inputting the N first feature sets corresponding to the N sub-images into the prediction module to obtain the chromosome detection result.
Optionally, the apparatus further comprises:
a generating module, configured to generate an initial homologous chromosome pairing result according to the chromosome detection result, where the chromosome detection result includes morphological information of chromosomes, and the initial homologous chromosome pairing result includes an initial number of each chromosome;
an obtaining input module, configured to obtain first prior data corresponding to each chromosome, and input the first chromosome into an alternative set when there is a first chromosome whose morphological information does not match the corresponding first prior data, where the first prior data is prior data matching an initial number of the corresponding chromosome;
an obtaining and updating module, configured to obtain at least one second priori data corresponding to each second chromosome, respectively, and if a third chromosome exists in all the second chromosomes, where morphological information of the third chromosome matches any second a priori data in the corresponding at least one second a priori data, update an initial number of the third chromosome to a number corresponding to the any second a priori data, where the second chromosome is a chromosome other than the first chromosome in all chromosomes in the chromosome detection result, and the second a priori data is a priori data that matches an adjacent number of the initial number of the corresponding chromosome;
an obtaining and determining module, configured to determine all target numbers with a number of corresponding chromosomes less than 2 from the updated initial numbers, obtain third prior data corresponding to each target number, and determine a number of the first chromosome in the candidate set according to the third prior data.
The chromosome detection apparatus is an electronic device corresponding to the chromosome detection method, and all the implementation manners in the method embodiments are applicable to the embodiment of the apparatus, and the same technical effects can be achieved.
Optionally, an embodiment of the present invention further provides an electronic device, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor implements the chromosome detection method when executing the computer program.
Optionally, an embodiment of the present invention further provides a computer-readable storage medium, where a computer program is stored, and when the computer program is executed by a processor, the method for detecting chromosomes is implemented.
The above-mentioned embodiments are only used for illustrating the technical solutions of the present application, and not for limiting the same; although the present application has been described in detail with reference to the foregoing embodiments, it should be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; such modifications and substitutions do not substantially depart from the spirit and scope of the embodiments of the present application and are intended to be included within the scope of the present application.
Claims (10)
1. A method for detecting a chromosome, comprising:
dividing the acquired chromosome image into N sub-images, wherein N is an integer greater than 1;
inputting the N sub-images into a chromosome detection model to obtain a chromosome detection result, wherein the chromosome detection model comprises a prediction module and N feature extraction modules, the N feature extraction modules correspond to the N sub-images one by one, the input ends of the N feature extraction modules are used for receiving the N sub-images, the output ends of the N feature extraction modules are connected to the input end of the prediction module, and the output end of the prediction module is used for outputting the chromosome detection result.
2. The method of claim 1, wherein the prediction module comprises a karyotype prediction unit, and the chromosome detection result comprises a chromosome karyotype;
inputting the N sub-images into a chromosome detection model to obtain a chromosome detection result, including:
inputting the N sub-image inputs into the N feature extraction modules correspondingly to obtain N feature sets;
and inputting the N feature sets into the karyotype prediction unit to obtain the chromosome karyotype.
3. The method of claim 2, wherein inputting the N feature sets into the karyotype prediction unit to obtain the chromosome karyotype comprises:
predicting each feature set in the N feature sets by the karyotype prediction unit to obtain N initial prediction results;
and calculating by the karyotype prediction unit according to the N initial karyotype prediction results and the preset weight of each initial karyotype prediction result to obtain the chromosome karyotype.
4. The method of claim 2, wherein the prediction module further comprises a polarity prediction unit, wherein the chromosome detection result further comprises a chromosome polarity;
after the N sub-image inputs are respectively and correspondingly input to the N feature extraction modules to obtain N feature sets, the method further includes:
and inputting the N feature sets into the polarity prediction unit to obtain the polarity of the chromosome.
5. The method of claim 4, wherein each sub-image of the N sub-images has a corresponding image number;
inputting the N feature sets into the polarity prediction unit to obtain the polarity of the chromosome, wherein the method comprises the following steps:
merging the N feature sets according to the image sequence number of the sub-image corresponding to each feature set through the polarity prediction unit to obtain a merging result;
and predicting by the polarity prediction unit based on the combined result to obtain the chromosome polarity.
6. The method according to claim 1, wherein the inputting the N sub-images into a chromosome detection model to obtain chromosome detection results comprises:
performing convolution processing on a first sub-image through a first feature extraction module to obtain a first convolution image, wherein the first feature extraction module is any one of the N feature extraction modules, and the first sub-image is a sub-image corresponding to the first feature extraction module in the N sub-images;
randomly generating a first mask through the first feature extraction module, and generating a first processed image according to the first volume image and the first mask;
extracting features based on the first processed image through the first feature extraction module to obtain a first feature set;
inputting the N first feature sets corresponding to the N sub-images into the prediction module to obtain the chromosome detection result.
7. The method according to claim 1, wherein after inputting the N sub-images into a chromosome detection model and obtaining chromosome detection results, the method further comprises:
generating an initial homologous chromosome pairing result according to the chromosome detection result, wherein the chromosome detection result comprises morphological information of chromosomes, and the initial homologous chromosome pairing result comprises an initial number of each chromosome;
respectively acquiring first prior data corresponding to each chromosome, and inputting the first chromosome into an alternative set under the condition that the first chromosome with form information not matched with the corresponding first prior data exists, wherein the first prior data is prior data matched with the initial number of the corresponding chromosome;
respectively obtaining at least one second prior data corresponding to each second chromosome, and updating an initial number of a third chromosome to a number corresponding to any second prior data when the third chromosome exists in all the second chromosomes, wherein the morphological information of the third chromosome is matched with any second prior data in the corresponding at least one second prior data, the second chromosome is a chromosome in all chromosomes in the chromosome detection results except the first chromosome, and the second prior data is prior data matched with a number adjacent to the initial number of the corresponding chromosome;
determining all target numbers with the number of corresponding chromosomes being less than 2 from the updated initial numbers, obtaining third prior data corresponding to each target number, and determining the number of the first chromosome in the alternative set according to the third prior data.
8. A chromosome testing device, comprising:
the segmentation module is used for segmenting the acquired chromosome image into N sub-images, wherein N is an integer greater than 1;
the detection module is used for inputting the N sub-images into a chromosome detection model to obtain a chromosome detection result, wherein the chromosome detection model comprises a prediction module and N feature extraction modules, the N feature extraction modules correspond to the N sub-images one by one, the input ends of the N feature extraction modules are used for receiving the N sub-images, the output ends of the N feature extraction modules are connected to the input end of the prediction module, and the output end of the prediction module is used for outputting the chromosome detection result.
9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the method of any of claims 1 to 7 when executing the computer program.
10. A computer-readable storage medium, in which a computer program is stored which, when being executed by a processor, carries out the method according to any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010774367.4A CN112037173B (en) | 2020-08-04 | 2020-08-04 | Chromosome detection method and device and electronic equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010774367.4A CN112037173B (en) | 2020-08-04 | 2020-08-04 | Chromosome detection method and device and electronic equipment |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112037173A true CN112037173A (en) | 2020-12-04 |
| CN112037173B CN112037173B (en) | 2024-04-05 |
Family
ID=73581926
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010774367.4A Active CN112037173B (en) | 2020-08-04 | 2020-08-04 | Chromosome detection method and device and electronic equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112037173B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112711983A (en) * | 2020-12-08 | 2021-04-27 | 湖南自兴智慧医疗科技有限公司 | Nuclear pattern analysis system, method, electronic device and readable storage medium |
| CN115330603A (en) * | 2022-10-17 | 2022-11-11 | 湖南自兴智慧医疗科技有限公司 | Human chromosome image rectification method based on deep learning convolutional neural network |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108038435A (en) * | 2017-12-04 | 2018-05-15 | 中山大学 | A kind of feature extraction and method for tracking target based on convolutional neural networks |
| CN109754391A (en) * | 2018-12-18 | 2019-05-14 | 北京爱奇艺科技有限公司 | A kind of image quality evaluating method, device and electronic equipment |
| CN110390676A (en) * | 2019-07-26 | 2019-10-29 | 腾讯科技（深圳）有限公司 | The cell detection method of medicine dye image, intelligent microscope system under microscope |
| CN110533684A (en) * | 2019-08-22 | 2019-12-03 | 杭州德适生物科技有限公司 | A kind of karyotype image cutting method |
| CN110533672A (en) * | 2019-08-22 | 2019-12-03 | 杭州德适生物科技有限公司 | A kind of chromosome sort method based on band identification |
| CN110555390A (en) * | 2019-08-09 | 2019-12-10 | 厦门市美亚柏科信息股份有限公司 | pedestrian re-identification method, device and medium based on semi-supervised training mode |
| WO2020042704A1 (en) * | 2018-08-27 | 2020-03-05 | 杭州德适生物科技有限公司 | Chromosome identifying method based on deep learning |
| CN111325739A (en) * | 2020-02-28 | 2020-06-23 | 北京推想科技有限公司 | Method and device for detecting lung focus and training method of image detection model |
-
2020
- 2020-08-04 CN CN202010774367.4A patent/CN112037173B/en active Active
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108038435A (en) * | 2017-12-04 | 2018-05-15 | 中山大学 | A kind of feature extraction and method for tracking target based on convolutional neural networks |
| WO2020042704A1 (en) * | 2018-08-27 | 2020-03-05 | 杭州德适生物科技有限公司 | Chromosome identifying method based on deep learning |
| CN109754391A (en) * | 2018-12-18 | 2019-05-14 | 北京爱奇艺科技有限公司 | A kind of image quality evaluating method, device and electronic equipment |
| CN110390676A (en) * | 2019-07-26 | 2019-10-29 | 腾讯科技（深圳）有限公司 | The cell detection method of medicine dye image, intelligent microscope system under microscope |
| CN110555390A (en) * | 2019-08-09 | 2019-12-10 | 厦门市美亚柏科信息股份有限公司 | pedestrian re-identification method, device and medium based on semi-supervised training mode |
| CN110533684A (en) * | 2019-08-22 | 2019-12-03 | 杭州德适生物科技有限公司 | A kind of karyotype image cutting method |
| CN110533672A (en) * | 2019-08-22 | 2019-12-03 | 杭州德适生物科技有限公司 | A kind of chromosome sort method based on band identification |
| CN111325739A (en) * | 2020-02-28 | 2020-06-23 | 北京推想科技有限公司 | Method and device for detecting lung focus and training method of image detection model |
Non-Patent Citations (2)
| Title |
|---|
| FAROUDJA ABID & LATIFA HAMAMI: "A survey of neural network based automated systems for human chromosome classification", ARTIFICIAL INTELLIGENCE REVIEW * |
| 张成成;宋婕萍;徐淑琴;李卉;徐闰红;王小艳;张立;游齐靖;张凯;林浩添;: "基于深度卷积神经网络对中期染色体分类的应用研究", 中国临床新医学, vol. 13, no. 02, pages 123 - 126 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112711983A (en) * | 2020-12-08 | 2021-04-27 | 湖南自兴智慧医疗科技有限公司 | Nuclear pattern analysis system, method, electronic device and readable storage medium |
| CN115330603A (en) * | 2022-10-17 | 2022-11-11 | 湖南自兴智慧医疗科技有限公司 | Human chromosome image rectification method based on deep learning convolutional neural network |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112037173B (en) | 2024-04-05 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109241903B (en) | Sample data cleaning method, device, computer equipment and storage medium | |
| CN109740534B (en) | Image processing method, device and processing equipment | |
| CN109063742B (en) | Butterfly identification network construction method and device, computer equipment and storage medium | |
| CN111814741B (en) | Method for detecting embryo-sheltered pronucleus and blastomere based on attention mechanism | |
| US20200210786A1 (en) | Computer-executed method and apparatus for assessing vehicle damage | |
| WO2022012110A1 (en) | Method and system for recognizing cells in embryo light microscope image, and device and storage medium | |
| CN109993102B (en) | Similar face retrieval method, device and storage medium | |
| CN111783779B (en) | Image processing method, apparatus and computer readable storage medium | |
| CN113222149A (en) | Model training method, device, equipment and storage medium | |
| CN112037173B (en) | Chromosome detection method and device and electronic equipment | |
| CN115424101A (en) | Disease identification method, device, equipment and storage medium | |
| CN111985488B (en) | Target detection segmentation method and system based on offline Gaussian model | |
| TW202011266A (en) | Neural network system for image matching and location determination, method, and device | |
| CN113780145A (en) | Sperm morphology detection method, sperm morphology detection device, computer equipment and storage medium | |
| CN114419313A (en) | Image identification method and image identification system | |
| CN116229066A (en) | Portrait segmentation model training method and related device | |
| CN115496976A (en) | Visual processing method, device, equipment and medium for multi-source heterogeneous data fusion | |
| CN115511798A (en) | Pneumonia classification method and device based on artificial intelligence technology | |
| CN117372604A (en) | 3D face model generation method, device, equipment and readable storage medium | |
| CN112001285A (en) | Method, device, terminal and medium for processing beautifying image | |
| CN111340025A (en) | Character recognition method, character recognition device, computer equipment and computer-readable storage medium | |
| CN110992371B (en) | Portrait segmentation method and device based on priori information and electronic equipment | |
| CN114494441B (en) | Grape and picking point synchronous identification and positioning method and device based on deep learning | |
| CN116310899A (en) | YOLOv 5-based improved target detection method and device and training method | |
| CN115760775A (en) | Tomato disease detection method, device, equipment and storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |